time series segmentation

ts_seg.py

import numpy as np

def slidingWindow(x,max_error):
	n=len(x)
	leftNode=0
	segmentList=[]
	print n
	
	while leftNode<n-1:
		print leftNode
		newSeg = False
		for rightNode in range(leftNode+3,n):
			testSeg=x[leftNode:rightNode]

			testLine = createLine(testSeg,"regression")

			segError = calculate_error(testSeg, testLine)

			if segError>max_error:
				segmentList.append([testLine,range(leftNode,rightNode)])
				leftNode=rightNode
				newSeg = True
				break

		if newSeg is False:
			segmentList.append([testLine,range(leftNode,rightNode)])
			leftNode = n-1

	return segmentList


def calculate_error(x,y):
	return np.max(np.abs(x-y))


def createLine(x,method="simple"):
	n=len(x)
	if method=="simple":
		# Using two node as line
		line = np.linspace(x[0],x[-1],n)
		pass

	elif method == "regression":
		# Using other node
		slope, intercept, r_value, p_value, std_err = ss.linregress(range(n),x)
		line = intercept+slope*np.arange(n)
	
	elif method == "poly":
		line=np.poly1d(np.polyfit(range(n),x,2))(range(n))

	return(line)